[Top] [All Lists]

Re: another 4kc machine check.

To: "Kevin D. Kissell" <>
Subject: Re: another 4kc machine check.
From: Greg Weeks <>
Date: Tue, 12 Apr 2005 12:52:18 -0400
Cc: "Maciej W. Rozycki" <>, Ralf Baechle <>,
In-reply-to: <00c701c53f7e$09ec56c0$10eca8c0@grendel>
Original-recipient: rfc822;
References: <> <> <> <> <> <004a01c53ed4$dab12b00$10eca8c0@grendel> <> <00c701c53f7e$09ec56c0$10eca8c0@grendel>
User-agent: Mozilla Thunderbird 1.0 (X11/20041206)
Kevin D. Kissell wrote:

On Mon, 11 Apr 2005, Kevin D. Kissell wrote:

If the 4KC and 4KEC need it, so does the 4KSC (and 4KSD).
But that's weird in the first place as 4Kc implements the original revision of MIPS32 so it does not implement "ehb". Therefore it acts just as an ordinary "nop", but according to the 4K manual there is no need for one -- the hazard between a move to EntryLo0/EntryLo1 and tlbwi/tlbwr is explicitly listed as 0 instructions.

Oops.  Maybe I misread the patch.  I thought the added NOP was between
the TLBWR and the ERET.
No, it adds it before the TLBWR where there shouldn't be a hazard.

Greg Weeks

<Prev in Thread] Current Thread [Next in Thread>